RockstarChatbots presented chatbots for business at a recent Catalyst88 conference for Amazon sellers. We were welcomed by the group and everyone helped to find uses for the bots. Being surrounded by business entrepreneurs is a great way to expose the finer details in deploying chatbots. Each chatbot should have its own unique tool set which aligns with user needs. Knowing what tools are available is one of the key elements in designing chatbots for business. Our presentation sought to define the chatbots in two ways: What their functionality is and where they can be deployed.
It helps to think of these functions and sub-functions as interchangeable. Any of them can be combined with another to drive the chatbot towards your chosen goal. However, each function has is good side and bad. Try to think of what your use case may be as we go through some of the variations and combinations.
Logic bots and their pre-determined responses work well for tech support or for equipment manuals. They may also work well in a sales funnel that drives all of its visitors to well defined outcomes. Conversely, they may fail when tasked with small talk and other chatter that an AI bot can be prepared for. An e-commerce site with only a few options could be handled by a relatively small question/answer set (logic bot). A large scale shopping site would probably need some advanced capabilities (AI or interactive).
One way to picture the differences is deciding who we would want to talk to if we were at a party. If we just need to know where the bathroom is or if the host is recycling, a logic bot would be fine. If we were sitting on the couch, the interactive bot may be the best at turning the conversation a pleasant way and if their were party games, the AI bot would be the best team mate.
Looking at it this way, we can see a product launch may use an AI chatbot to project its personality and engage users. Later the scope could be tightened to more effectively define the product and drive users to the buy button. An interactive bot could be tweaked to provide customers with the proper information, help make choices and walk them through to purchase. Once the customer needs and the product attributes are defined, the job becomes easier and a logic bot could pass out technical info and how-to’s; handle returns and decide who gets coupons.
Most DIY and provider chatbots will use a combination of these functions to get a desired result. An important consideration for all of us is cost. Predictably, the more functions and AI a chatbot uses, the higher the cost will be. However, providers such as Chatfuel or ManyChat enable users to build bots for free. We can build bots quickly on these platforms using our brains, good faith and humor. Some of the options are easy to use and they can be built into powerful AI bots with IBM’s Watson.
Messenger chatbots are built with the Chatfuel application. It is a powerful tool that integrates with Facebook’s many communicative and advertising functions. Messenger chatbots make great assistants when running a major Facebook ad campaign or simply engaging users with rapid responses on any Facebook business site. Messenger bots are currently the favorite children in the fledgling chatbot industry. Their only drawback is that they can only be deployed on Facebook sites by Facebook users.
Enter RockstarChatbots’ Weber. We built the Web Chatbot and it’s UI Botbuilder to break through the constraints posed by many of the available bot platforms. We wanted a lightweight chatbot that could be deployed anywhere on the web. Additionally, we realized a great variety of user needs would require greater flexibility. Weber is hand crafted on baseline computer languages which can mitigate problems caused by dependence on other programs or applications. The web chatbot’s stability and flexibility enhance it’s charming geniality. We believe it will become a crowning achievement in bot flexibility and extensibility.
The AI Chatbot is simply the Web Chatbot supercharged with any applications the web has to offer. Artificial Intelligence, Natural Language Processing (NLP), machine learning (ML) and huge server arrays are combined to produce a cognitive experience. This Lamborghini of the bot world can literally think for itself. These types of bots are making the news; extremely capable in all sorts of human-like behavior. Our proprietary algorithms focus their abilities and keep them working for you.
As usual this came to me in the middle of the night. Well, not really. Last night I was working diligently on an un-related problem. This popped into my head when I woke up this morning.
We pay about 30 bucks a month for our home security system. It monitors entry points such as doors and windows while we are away. It also monitors movement inside the house. If there is a break-in or some other emergency the system goes to work. A cascade of events “decides” if it really is a burglary or fire and only then calls the proper authority. There is also a panic button similar those Flava-flave necklaces we see on TV; “I’ve fallen and I can’t get up!”
It’s all good but the problem stems from the cost. We pay too much and there are never emergencies. A chatbot seems like the perfect solution.
There are a few other steps involved but through this example we can see how chatbots can provide emergency services with a clear picture of the situation. Security is much improved because we don’t have to hand over information about ourselves unless there is a real need for it. The police, fire and ambulance services should love this; a virtual picture to help them decide on their tactics. A chatbot armed with scripts used by alarm companies and 911 operators could vastly improve the current status quo at huge cost savings for all. We just need to get around to it.
It has become a standard: A robot took my job. This mantra should be familiar to everyone as mechanization in the workplace has been an issue for at least one hundred years. In 1913, Henry Ford’s automated assembly line turned a whole company and its people into a giant cyborg; half human and half robot. We could further trace robot outsourcing back through mid 19th century British industrialists, scientists and free thinkers who spent their time developing machines that “replaced” people. If we consider Archimedes screw that “laid-off” people who were moving water with buckets it reaches the edge of written history.
We spend half our time vilifying people like Henry Ford and the other half celebrating them. However, we should probably take a few moments every Friday afternoon to thank Ford for inventing the weekend and an 8 hour work day. Who really knows if his motives were altruistic or motivated purely by greed in optimizing his manufacturing plant. We do know he created a car that anyone could afford when only rich people could afford cars. We also know he was the first to institutionalize the concept of a life-work balance.
Thinking through this, there is reason to believe that robots are all-good for people. Consider the possibility that jobs we are “losing” to robots are really just parts of jobs that are too tedious or back breaking to do for 8 or 10 hours every day. Let’s face it, the good parts of any job are the parts that require some imagination or intuitive thinking. Why not think of any job or task as part that a robot can do and part that needs real thought and flexibility?
If we start to draw this out we can see how a robot taking part of a job may benefit the employer, the worker and the client or end user. Since this is a chatbot blog let’s consider a product support chat room that services 100 calls a day with 4 people on the team. Each product specialist handles an average of 25 calls a day. It doesn’t really matter what the numbers are. We just want to be able to see the effects of automation.
A pie chart will be good to handle this:
Maybe it would be better to add some humanity. People don’t work 100% of the time so let’s add a simple buffer that may include getting a quick snack, adjusting the work station, maintenance, going to the toilet or taking a deep breath once in a while. In a way, this number is a constant for any amount of humans. Every person needs humanity in their job and there will be a fitful price to be paid if it is not applied. We’ll use 10%. It is probably more like 30% but let’s use 10% to avoid arguments at this point.
Now let’s add a robot into the mix. Consider how many calls the product specialists will receive that have existing answers. They are either in the frequently asked questions (FAQ’s) for the given product or a chat has produced a question and answer that can be used in the FAQ’s. Assuming these FAQ’s and their answers are not changing, we can give them to a chatbot. The chatbot can filter incoming requests and hand-off the customer when it can’t find a solution. It is important to note that the chatbot has to handle this gracefully or it will just be giving annoyed customers to the specialists. We’ll use 15% as the percentage of requests a chatbot may handle.
This is the part where a lean thinking manager might say “Aha! Now I can get rid of one of my support agents if that chatbot will just handle a few more calls.” This could be a good idea. It is entirely possible a chatbot could buffer 80-90% of increased support calls during an upsurge from product deployments or failures. Conversely, this might not be a good idea. What if there is a growth spurt in the company or a new feature or product is launched…and fails? A chatbot can easily handle repeat questions such as “How do I return this?” or “Where’s the manual?” but a good support tech may do better at leveraging nuances in the conversation to endear good faith from the customer. In this case, a chatbot combined with the company’s loyal support people may outperform an outsourced team dedicated to events such as recalls.
A key point is: the chatbot is “in-sourced.” An in-house chatbot has the advantage of taking input from the sales team along with support or engineering. The well mannered chatbot can appease the customer while offering references to new products and public relations based media. We could give the support team a script produced by the sales team along with some training to solve the problem. On the other hand, that would take the support team out of their element while creating more administrative and training work. Arguably, a well tuned chatbot with a good script will cost less and perform better than outsourced or overworked support teams.
Business owners and managers will want to know about the bottom line. The US Bureau of Labor Statistics lists the annual mean salary for office and administrative support occupations at $36,000 per year. Considering costs for administration such as payroll and HR, fees, insurance and taxes, the total cost to hire support people is around twice their salary: $72,000 per year. A manager may be looking at this when increasing the size of the support team. To the manager a chatbot might be looking pretty good right now.
At this point we may have the managers ear but the “robot took my job” argument has come to the forefront. Let’s see if we can make everybody happy…
First, the manager since the ball is already rolling. Looking at the pie chart with the chatbot we can see the chatbot’s job portion is very close to the human worker’s part. If the given chatbot was optimized a bit more it could effectively “replace” the fifth human on the team. Also, remember the chatbot’s ability to simultaneously handle multiple calls and work all three shifts. If we can ensure the chatbot won’t mishandle customers, the manager is happy.
The support team may balk at implementing a chatbot that could take their job. A suitable device will be needed to convince them. Let’s use a carrot! The key to any job is growth. We need to get a raise once in a while to keep up with inflation or a growing family. We may also get bored with a given post and want a way up…or out. A training program that involves learning about and tuning the chatbot may provide a perfect opportunity for the worker to gain knowledge and skills and for the company to improve employee retention.
One support person might find chatbot programming interesting and take night classes to get a better job in the company. Another might go off to school and return with diploma in hand, thankful for the company’s encouragement and ready to make a positive difference. Still another may be sad at loosing the easy questions to the chatbot but somewhat satisfied the manager kicked down a few bucks at his last review. It was all possible because of the improvements in service created by the chatbot and training program. The team is mostly happy. The two new members are very happy because they found jobs when one member was promoted and the other went to college.
And finally, the customer. Why last? Because customer satisfaction and conversion is the bottom line in any business. How we treat the customer with courtesy and respect while we are cajoling, coaxing and coercing them to like our products and buy more will advance our chances for success in business. The customer may not always be right but customers must be listened to and the knowledge gained must be employed effectively. By using chatbots to courteously interact with our customers we have an opportunity to streamline support issues, increase leads and sales, and mine for data. The customer gets treated carefully and fairly; gets the product they are looking for; and enjoys time spent on your website. The customer is happy.
One thing I learned as a computer engineering student (and as my father’s son) is that a very important part of engineering is clarifying what a product does not do. In the engineer’s risk-centric world, avoiding risks may be more important than what a product really does. So much of the time developing or designing a product we ask ourselves “What is this thing supposed to do?” We consult with our clients, making sure not to miss a feature or special function. It is usually when a product does something or does something so horribly that we find ourselves called in on the carpet explaining what went wrong.
If nothing at all happens when we hit the start button then that elicits the Hippocratic method; Above all do no harm. If the car starts and drives through the front of a Quickie-Mart then the Hippocratic car rules have just been smashed along with all the sodas, chips and cans of Dinty Moore stew. This set of examples drives the engineering and business field of risk-analysis. Software developers operate under the same guidelines. However, in software such as web development a lot of risk assessment is seen as resistance to getting something on the screen which is generating leads and converting visitors into customers.
Support bots and virtual assistants operate under the same rules as everyone else. They must weigh the advantages of a risk assessment strategy along with the cost of producing one and seeing it through. The difference for chatbots is that they may be performing the tech support duties along with the lead generation and sales. A chatbot can act as a singular collaborative team with perfect communication skills balancing the requirements in real time. We have all heard of a marketer promising the world and the production team scrambling like mad to make something that resembles the incredible promise. The chatbot working both ends of a deal may think twice before overly bragging. Additionally, the well programmed chatbot can be made to consider quotas and return on investment (ROI) metrics while it is effortlessly chatting up a customer and flashing photos of kittens and puppies.
As we can see, chatbots can take an integral role in balancing risk aversion tactics with the need to convert sales and pay the rent. We will always need people to make the rules and deeply consider what is best for our business and clients. Luckily, we will soon be embracing bots for their remarkable ability to see the big picture and play by the rules. The added plus is: It will be safer to loiter in the magazine aisle in Quickie-Marts.